Bugzilla Bug 193513

With x11-libs/gtk+-2.12.0 and dev-libs/glib-2.14.1, new funcitonality
introduced in glib causes nspluginviewer when loaded in a non-GTK environment
(KDE in my case) to hang, since flash never initializes GTK properly. According
to a bug report on Gnome's bugzilla, this also causes problems for Openoffice.
Patch to follow.

Reproducible: Always

Steps to Reproduce:
1. Emerge glib-2.14.1 and gtk+-2.12.0 in a non-GTK environment
2. Go to youtube.com
3. Watch the the CPU go to 100%



Portage 2.1.3.9 (default-linux/x86/2007.0, gcc-4.2.0, glibc-2.6.1-r0,
2.6.22-gentoo-r6 i686)
=================================================================
System uname: 2.6.22-gentoo-r6 i686 AMD Sempron(tm)
Timestamp of tree: Sun, 23 Sep 2007 04:50:01 +0000
ccache version 2.4 [enabled]
app-shells/bash:     3.2_p17-r1
dev-java/java-config: 1.3.7, 2.0.33-r1
dev-lang/python:     2.5.1-r2
dev-util/ccache:     2.4-r7
sys-apps/baselayout: 2.0.0_rc4-r1
sys-apps/sandbox:    1.2.18.1
sys-devel/autoconf:  2.13, 2.61-r1
sys-devel/automake:  1.5, 1.7.9-r1, 1.8.5-r3, 1.9.6-r2, 1.10
sys-devel/binutils:  2.18
sys-devel/gcc-config: 1.4.0-r2
sys-devel/libtool:   1.5.24
virtual/os-headers:  2.6.22-r2
ACCEPT_KEYWORDS="x86 ~x86"
CBUILD="i686-pc-linux-gnu"
CFLAGS="-O2 -march=athlon-xp -pipe -ggdb3"
CHOST="i686-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/kde/3.5/env /usr/kde/3.5/share/config
/usr/kde/3.5/shutdown /usr/share/X11/xkb /usr/share/config"
CONFIG_PROTECT_MASK="/etc/env.d /etc/env.d/java/ /etc/fonts/fonts.conf
/etc/gconf /etc/gentoo-release /etc/revdep-rebuild /etc/splash /etc/terminfo
/etc/udev/rules.d"
CXXFLAGS="-O2 -march=athlon-xp -pipe -ggdb3"
DISTDIR="/usr/portage/distfiles"
FEATURES="buildpkg ccache distlocks metadata-transfer parallel-fetch sandbox
sfperms splitdebug strict unmerge-orphans userfetch"
GENTOO_MIRRORS="http://ftp.belnet.be/mirror/rsync.gentoo.org/gentoo/
http://pandemonium.tiscali.de/pub/gentoo/ http://mirror.gentoo.no/
http://gentoo.osuosl.org/"
LANG="en_US.UTF-8"
LC_ALL="en_US.UTF-8"
LINGUAS="en da"
MAKEOPTS="-j2"
PKGDIR="/usr/portage/packages"
PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress
--force --whole-file --delete --delete-after --stats --timeout=180
--exclude=/distfiles --exclude=/local --exclude=/packages
--filter=H_**/files/digest-*"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/local/portage"
SYNC="rsync://rsync.europe.gentoo.org/gentoo-portage"
USE="3dnow 3dnowext X a52 aac acpi aim alsa authdaemond bash-completion berkdb
bitmap-fonts cairo cdr cjk cli cpudetection cracklib crypt css cups curl dbus
djbfft dri dts dvd dvdr dvdread evdev fam ffmpeg firefox flac fortran g15 gadu
gif glitz gnutls gpm hal hbci iconv icq idea imap isdnlog jabber java
javascript jpeg kde kdeenablefinal lcd lm_sensors logitech-mouse logrotate lzo
mad matroska mbox midi mikmod mime mmx mmxext mono moznopango mp3 mp4 mpeg
mplayer msn mudflap musepack ncurses network-cron nls nocd nptl nptlonly
offensive ogg openal opengl openmp oscar pam pcre pdf perl png pop ppds pppd
python qt3 qt4 quicktime readline reflection sasl sdl sensord session shorten
slang smime sms speex spell spl sse ssl svg svga tcpd texteffect theora threads
tiff translator truetype truetype-fonts type1-fonts unicode usb vcd vorbis
win32codecs wmf wxwindows x264 x86 xanim xcb xface xine xml xorg xv xvid yahoo
zlib" ALSA_CARDS="via82xx" ALSA_PCM_PLUGINS="adpcm alaw asym copy dmix dshare
dsnoop empty extplug file hooks iec958 ioplug ladspa lfloat linear meter mulaw
multi null plug rate route share shm softvol" ELIBC="glibc"
INPUT_DEVICES="keyboard mouse evdev" KERNEL="linux" LCD_DEVICES="g15"
LINGUAS="en da" USERLAND="GNU" VIDEO_CARDS="fglrx vesa radeon"
Unset:  CTARGET, EMERGE_DEFAULT_OPTS, INSTALL_MASK, LDFLAGS, PORTAGE_COMPRESS,
PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS

Created an attachment (id=131686) [details]
gtk+-2.12.0.patch

Patch for gtk+-2.12.0.ebuild
See:
http://www.archlinux.org/news/338/
http://cvs.archlinux.org/cgi-bin/viewcvs.cgi/lib/gtk2/?cvsroot=Extra&only_with_tag=CURRENT
https://bugzilla.novell.com/show_bug.cgi?id=294385

Created an attachment (id=131690) [details]
gtkclipboard-check.patch

To be placed in ${FILESDIR}/gtkclipboard-check.patch

Created an attachment (id=131691) [details]
gdkscreen-check.patch

To be placed in ${FILESDIR}/gdkscreen-check.patch

This ebuild with patches solves the problem for me. Taken from Arch's cvs. You
may want to mark this as a blocker of BUG 193369.

I confirm both the issue and the patches solving.

It may as well be a blocker for bug #182561 (likely) and for
the reopened bug #177446 (less likely).

I use konqueror. It hangs when getting on a page using flash. The patches fix
this issue.

I have included a proper patch for the clipboard in gtk+-2.12.0-r1 that was
attached to the GNOME upstream bug report (not novell, nor arch linux). Thanks
for the links, they lead me to the upstream bug :)  Testing of the revbump
would be appreciated.
What is the need for the gdkscreen-check patch? What does it do/fix, why,
where's the backtrace or upstream bug?
Until there is a known and understood reason I personally will not include it

(In reply to comment #7)
> I have included a proper patch for the clipboard in gtk+-2.12.0-r1 that was
> attached to the GNOME upstream bug report (not novell, nor arch linux). Thanks
> for the links, they lead me to the upstream bug :)

You could have just, you know, looked at the URL of this bug...

>  Testing of the revbump
> would be appreciated.

WFM with flash

> What is the need for the gdkscreen-check patch? What does it do/fix, why,
> where's the backtrace or upstream bug?

That's the fix for the openoffice bug:
http://www.openoffice.org/issues/show_bug.cgi?id=80823
OOo refers that bug to Ubuntu, though it's really their software that fails to
initialize GTK+ properly.

> Until there is a known and understood reason I personally will not include it

There.

(In reply to comment #8)
> You could have just, you know, looked at the URL of this bug...

These should be more prominent :) Sorry

> > What is the need for the gdkscreen-check patch? What does it do/fix, why,
> > where's the backtrace or upstream bug?
> 
> That's the fix for the openoffice bug:
> http://www.openoffice.org/issues/show_bug.cgi?id=80823
> OOo refers that bug to Ubuntu, though it's really their software that fails to
> initialize GTK+ properly.

> > Until there is a known and understood reason I personally will not include it
> 
> There.

I didn't find a reasoning or backtrace from any of the links provided. Maybe
I'm continuously blind. Furthermore, openoffice-bin works great for me so I
can't test (I have x86 btw). Maybe OpenOffice.org bug should be reopened and
clarified? Misuse of gtk+ initialization is a bug for any platform, not just
their non-supported binary arches. Don't they even support the intialization
code itself, or what? :p

(In reply to comment #9)
> I didn't find a reasoning or backtrace from any of the links provided. Maybe
> I'm continuously blind. Furthermore, openoffice-bin works great for me so I
> can't test (I have x86 btw). Maybe OpenOffice.org bug should be reopened and
> clarified? Misuse of gtk+ initialization is a bug for any platform, not just
> their non-supported binary arches. Don't they even support the intialization
> code itself, or what? :p

Nope. Apparently this, like the Flash bug is only visible to a non-Gnome
environment, since Gnome initializes GTK+ for Openoffice. Or that's the way I
understand it. I can't confirm the bug since Openoffice isn't installed. I'm
for closing this bug and just letting the Openoffice thing be, unless someone
else can confirm it.
I did find a post on the forums where someone was experiencing the bug, though:
http://forums.gentoo.org/viewtopic.php?t=577955
AIUI, this person will get a hanging Openoffice when glib 2.14.1 goes stable.

I was thinking to include the patch in a slightly modified form as it doesn't
really hurt anything - there's an access to the argument in the next check
anyway with GDK_IS_SCREEN (not sure why that doesn't return false already when
screen == NULL). That when I have to do another revbump for other reasons soon
(simple search backend in filechooser crashing in certain situations), not
before.

The "modified form" means that I used
g_return_val_if_fail (screen != NULL, NULL) instead of "if ... then return
NULL"

I still hope to get some upstream input on this if it already isn't given on
some upstream bug we don't know about yet.

I can confirm the bug was there, repeatably, with gtk+-2.12.0.
Today, that is with gtk+-2.12.0-r1, nspluginviewer hasn't gone mad, yet.

I can't tell whether the bug has been really fixed.
I still get this in .xsession-errors, don't know if it's related:

(process:21593): Gtk-CRITICAL **: gtk_clipboard_get_for_display: assertion
`display != NULL' failed
Adobe FlashPlayer: gtk_clipboard_get(GDK_SELECTION_PRIMARY); failed. Trying to
call gtk_init(0,0);

If you need an environment to test with, mine seems to apply.

(In reply to comment #12)
> I can confirm the bug was there, repeatably, with gtk+-2.12.0.
> Today, that is with gtk+-2.12.0-r1, nspluginviewer hasn't gone mad, yet.

Yes, -r1 should fix the flash/nsplugin problem.

> I can't tell whether the bug has been really fixed.
> I still get this in .xsession-errors, don't know if it's related:
> 
> (process:21593): Gtk-CRITICAL **: gtk_clipboard_get_for_display: assertion
> `display != NULL' failed
> Adobe FlashPlayer: gtk_clipboard_get(GDK_SELECTION_PRIMARY); failed. Trying to
> call gtk_init(0,0);

Adobe FlashPlayer is doing stupidity by checking if gtk has been initialized
yet by calling gtk_clipboard_get_for_display to check that. However that is
wrong as gtk must be initted before this is called and this is assumed to be
done now and the README says so now - there are no guarantees from gtk+ that
things won't fail if such gtk functions are called before gtk_init is called.
So instead of calling gtk_clipboard_get_for_display to check if gtk_init has to
be called or not, Adobe FlashPlayer should just call gtk_init and that's it -
there are no ill effects if it's called even thousand times and never has been,
it will simply return without reiniting. So the assert comes because we patched
it to do so, to not simply fail but return and have it worked around. The
assert will disappear once it can disappear - that is when Adobe fixes it to
not do this "stupidity" :)

We are talking about the OpenOffice.org hang now, which a different patch
should help with (the gdkscreen-check.patch attached here) which isn't included
in portage yet.

> If you need an environment to test with, mine seems to apply.

Well, it's clear already that the gdkscreen-check patch works around the
problem, but my remaining concerns is "why is this workaround necessary and
should this be included by upstream as a fix or should openoffice be fixed
instead or both". I'm leaning towards one of the two former ones right now
(just openoffice fixed or both that and gtk+ workaround included in upstream,
but I will have to seek some upstream opinion on this).

gdkscreen-check (with the approach change to g_return_val_if_fail) for
openoffice in gtk+-2.12.0-r2 now together with a crash fix for the simple
search (without beagle nor tracker a `find` alike method) in filechooser when
closing the dialog before crawling ended.
I would like to keep this bug open for a bit until I or someone looks into this
at the upstream level. Basically I just lack the time for today for that, so
this bug will remind me to do it tomorrow.

Thanks a lot for the patch

CC'ing OOo maintainers for input.
Ok, I've tracked down some stuff. If my research is correct, this crash will
never happen with the openoffice-bin ebuilds. That's because the code that
caused this was introduced into many distributions through a common patchset.
The patch in question is patches/src680/ooo59127.vcl.honourcairofont.diff. The
old patch had a section of code like this:
+    if (const cairo_font_options_t *pOptions =
gdk_screen_get_font_options(gdk_screen_get_default()))
+       cairo_ft_font_options_substitute(pOptions, pPattern);

In the newer version (which is included in openoffice-2.3.0.ebuild), this is
the code:
+    GdkScreen *pScreen = gdk_screen_get_default();
+    if (const cairo_font_options_t *pOptions = pScreen ?
gdk_screen_get_font_options(pScreen) : 0)
+       cairo_ft_font_options_substitute(pOptions, pPattern);

And that fixes it. So, basically, this should no longer be a problem for
openoffice on Gentoo. The older ooo59127.vcl.honourcairofont.diff is included
in the openoffice-2.2.1-r1.ebuild, but could presumably trivially be changed to
the newer version. The only change between the two is that which is noted
above.

BTW,
Are my eyes deceiving me or did you guys bump 2.3.0 right into x86 stable?

(In reply to comment #15)
> And that fixes it. So, basically, this should no longer be a problem for
> openoffice on Gentoo. The older ooo59127.vcl.honourcairofont.diff is included
> in the openoffice-2.2.1-r1.ebuild, but could presumably trivially be changed to
> the newer version. The only change between the two is that which is noted
> above.

I would be happy to remove the workaround patch in a future version/revision of
gtk+ as soon as the openoffice team has a fixed version at best visibility
everywhere. Please notify me once this can be done

> BTW,
> Are my eyes deceiving me or did you guys bump 2.3.0 right into x86 stable?

revision 1.6
date: 2007-09-20 10:29:57 +0300;  author: opfer;  state: Exp;  lines: +2 -2; 
commitid: 6c1246f221734567;
stable x86, security bug 192818
(Portage version: 2.1.3.9)

Apparently without a ChangeLog entry.

I'm still experiencing problems with openoffice-bin and thunderbird when
x11-libs/gtk+-2.12.0-r2 that does not appears when I downgrade to gtk+-2.10.14.
It seems to me that it is related to this bug.

Specifically, here is what happens:
Thunderbird (version 2.0.0.6) crash when trying to write a new mail (the window
opens), the following message is shown in the terminal:
*** glibc detected *** /usr/lib/mozilla-thunderbird/thunderbird-bin:
munmap_chunk(): invalid pointer: 0x08f5f8a0 ***
/usr/libexec/mozilla-launcher: line 119:  9477 Aborted                 $(type
-P aoss) "$mozbin" "$@"
thunderbird-bin exited with non-zero status (134)

And Openoffice-bin (version 2.3.0) hangs when I try to open a new file through
the menu entry or when i try to export a document to PDF. The terminal shows:
*** glibc detected *** /usr/lib/openoffice/program/soffice.bin: free(): invalid
pointer: 0x08330730 ***

The only way I have found to solve this is to downgrade gtk+ and block the new
version.

Here is my emerge --info output:
Portage 2.1.3.11 (default-linux/x86/2007.0/desktop, gcc-4.2.0, glibc-2.6.1-r0,
2.6.22-suspend2-r2 i686)
=================================================================
System uname: 2.6.22-suspend2-r2 i686 Intel(R) Pentium(R) M processor 2.13GHz
Timestamp of tree: Mon, 01 Oct 2007 07:50:01 +0000
ccache version 2.4 [enabled]
app-shells/bash:     3.2_p17-r1
dev-java/java-config: 1.3.7, 2.0.33-r1
dev-lang/python:     2.4.4-r4, 2.5.1-r2
dev-python/pycrypto: 2.0.1-r6
dev-util/ccache:     2.4-r7
dev-util/confcache:  0.4.2-r1
sys-apps/baselayout: 1.12.10-r5
sys-apps/sandbox:    1.2.18.1
sys-devel/autoconf:  2.13, 2.61-r1
sys-devel/automake:  1.4_p6, 1.5, 1.6.3, 1.7.9-r1, 1.8.5-r3, 1.9.6-r2, 1.10
sys-devel/binutils:  2.18
sys-devel/gcc-config: 1.4.0-r2
sys-devel/libtool:   1.5.24
virtual/os-headers:  2.6.22-r2
ACCEPT_KEYWORDS="x86 ~x86"
CBUILD="i686-pc-linux-gnu"
CFLAGS="-march=pentium-m -O1 -pipe -fomit-frame-pointer"
CHOST="i686-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/kde/3.5/env /usr/kde/3.5/share/config
/usr/kde/3.5/shutdown /usr/share/config"
CONFIG_PROTECT_MASK="/etc/env.d /etc/env.d/java/ /etc/gconf
/etc/php/apache2-php5/ext-active/ /etc/php/cgi-php5/ext-active/
/etc/php/cli-php5/ext-active/ /etc/revdep-rebuild /etc/terminfo
/etc/texmf/web2c /etc/udev/rules.d"
CXXFLAGS="-march=pentium-m -O1 -pipe -fomit-frame-pointer
-fvisibility-inlines-hidden"
DISTDIR="/usr/portage/distfiles"
FEATURES="ccache distlocks metadata-transfer parallel-fetch sandbox sfperms
strict unmerge-orphans userfetch"
GENTOO_MIRRORS="http://sunsite.cnlab-switch.ch/mirror/gentoo
ftp://ftp.tu-clausthal.de/pub/linux/gentoo
ftp://sunsite.informatik.rwth-aachen.de/pub/Linux/gentoo
http://www.ibiblio.org/pub/Linux/distributions/gentoo"
LDFLAGS="-Wl,-O2 -Wl,--sort-common"
MAKEOPTS="-j1"
PKGDIR="/usr/portage/packages"
PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress
--force --whole-file --delete --delete-after --stats --timeout=180
--exclude=/distfiles --exclude=/local --exclude=/packages
--filter=H_**/files/digest-*"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/portage/local/layman/fluidportage
/usr/portage/local/layman/nx /usr/portage/local/layman/enlightenment
/usr/portage/local/layman/sunrise /usr/local/portage"
SYNC="rsync://rsync.de.gentoo.org/gentoo-portage"
USE="X a52 aac acl acpi acpi4linux aiglx alsa ansi asf avahi bash-completion
berkdb bitmap-fonts bluetooth bonjour boost bzip2 bzlib cairo cdr cli cracklib
crypt cups curl daap dba dbus dga divx divx4linux dlloader dri dts dvd dvdr
dvdread emacs emboss encode ethereal evo exif faad fam ffmpeg firefox flac
fortran gcj gd gdbm gecko-sdk gif gimp gimpprint glitz glut gmedia gmp gnome
gpm graphviz gstreamer gstreamer010 gtk gtk2 gtkhtml hal iconv id3 imap ipod
iproute2 isdnlog ithreads javascript jpeg jpeg2k kdeenablefinal
kdehiddenvisibility kerberos lcms libnotify live logitech-mouse logrotate lzw
mad maildir matroska mbox md5sum midi mikmod mime mjpeg mmap mmx mmx2 mmxext
mng mozcalendar moznoirc mozsvg mp3 mpeg mudflap musicbrainz ncurses network
nls nptl nptlonly ogg opengl openmp pam pcmcia pcre pdf perl pic plotutils png
ppds pppd prebuilt python qt-mt qt3 qt3support qt4 quicktime randr readline
real realmedia reflection reiserfs rtc samba sdl session simplexml smime
sockets speex spell spl sse sse2 ssl startup-notification stroke svg sysfs
t1lib tagwriting tcpd tetex theora threads tiff truetype truetype-fonts
type1-fonts unicode usb vcd vidix vorbis webdav win32codecs wmf wmp x264 x86
xcb xface xforms xine xinerama xml xml2 xorg xosd xpm xprint xscreensaver xv
xvid xvmc zeroconf zlib" ALSA_CARDS="intel8x0" ALSA_PCM_PLUGINS="adpcm alaw
asym copy dmix dshare dsnoop empty extplug file hooks iec958 ioplug ladspa
lfloat linear meter mulaw multi null plug rate route share shm softvol"
ELIBC="glibc" INPUT_DEVICES="evdev keyboard mouse synaptics void"
KERNEL="linux" LCD_DEVICES="bayrad cfontz cfontz633 glk hd44780 lb216 lcdm001
mtxorb ncurses text" USERLAND="GNU" VIDEO_CARDS="ati vesa radeon"
Unset:  CTARGET, EMERGE_DEFAULT_OPTS, INSTALL_MASK, LANG, LC_ALL, LINGUAS,
PORTAGE_COMPRESS, PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS

(In reply to comment #17)
> I'm still experiencing problems with openoffice-bin and thunderbird when
> x11-libs/gtk+-2.12.0-r2 that does not appears when I downgrade to gtk+-2.10.14.
> It seems to me that it is related to this bug.

No. This is not related to this bug, and any further communication about your
problem should be filed in a seperate bug. Before you file a new bug, please
try if your thunderbird problem can be solved with this:
emerge -1 mozilla-thunderbird
(you'll need to reinstall gtk+-2.12.0-r2)

HTH

For what it is worth, I'm seeing nspluginviewer crashes using
x11-libs/gtk+-2.12.1-r2 and net-www/netscape-flash-9.0.115.0 under kde-3.5.8
and konqueror.

openoffice, flash bugs should be fixed in gtk+-2.12.1-r2 and later on in
2.12.5-r1. If it's not please reopen this bug.

Concerning nspluginviewer, if it's still a bug with latest glib and gtk+,
please open a new bug to avoid confusion thanks.

(In reply to comment #20)
> openoffice, flash bugs should be fixed in gtk+-2.12.1-r2 and later on in
> 2.12.5-r1. If it's not please reopen this bug.
> 
> Concerning nspluginviewer, if it's still a bug with latest glib and gtk+,
> please open a new bug to avoid confusion thanks.
> 

After update to gtk+-2.12.10-r1 flash stop working in konqueror on site
live.palmtv.ru, youtube and other works fine. This problem is only in
konqueror, opera and firefox open and play video normaly.

(process:5293): Gtk-CRITICAL **: gtk_clipboard_get_for_display: assertion
`display != NULL' failed
Adobe Flash Player: gtk_clipboard_get(GDK_SELECTION_PRIMARY); failed. Trying to
call gtk_init(0,0);
Not initializing the Gtk-Qt theme engine

For reference: gtk+-2.12.0-openoffice-freeze-workaround.patch will not be
included in gtk+-2.14 series that I'm adding to portage in a bit, because it
seems that was a problem only for old openoffice versions which should be gone
by now. If problems occur, it can be restored.
The clipboard workaround (gtk+-2.12.0-flash-workaround.patch) will remain for
now.